To our knowledge, this is one of the few studies that assessed the prevalence of MM using a population-based sample and the first of its kind in Switzerland. Our results show that MM is relatively common in an apparently healthy general population. Our results also show that the prevalence of MM varies significantly according to the criteria used and even according to the data collection method (reported or measured).
Prevalence of multimorbidity
Prevalence of MM varied considerably according to the criteria used, a finding already reported in the literature [9,11], although this statement has been challenged [14]. Even using the same set of criteria, the prevalence varied considerably when self-reported or measured data was collected, a finding also noted when using data from electronic records or from health surveys [18,19]. A likely explanation is that many subjects are unaware of their status, as it has been shown for cardiovascular risk factors such as hypertension [20] or type 2 diabetes [21]. The lower prevalence of MM according to the FCI might also be due to the fact that the number of criteria is lower than the other definitions. Hence, a condition present in definitions A and B might not be considered as such with the FCI definition. Our results suggest that the prevalence of MM depends on the number of conditions considered, the higher the number the higher the likelihood of being diagnosed with MM. A possible (but not optimal) solution would be to modulate the threshold according to the number of criteria used to facilitate comparison between studies: for instance, MM could be diagnosed if a participant has 30% of all conditions, instead of a fixed number of conditions. Another possibility would be to select different definitions of MM according to the objective of the study [11], but this possibility would limit comparisons to studies with the same aims.
Although significant correlations were found between the number of reported or measured morbidities, still no good agreement was found between definitions, as only one third of participants diagnosed with MM by at least one definition was jointly diagnosed as MM by all three definitions. Our results thus stress the need for a common, standard definition of MM, which will allow comparison between studies.
The high prevalence of MM in our study also raises the question of the adequate management of subjects with MM. Indeed, health care providers are usually trained to manage one disease at the time (single-disease approach), with a further specialization for some conditions. Our results thus question the current status of medical training and of the medical system, and indicate that future medical education should bring back together different branches of medicine, as already postulated [22].
Determinants of multimorbidity
Women had higher rates of MM than men, and this association was confirmed after multivariate adjustment. These findings are in agreement with some studies [12,23] but not with others [2,14,24]. One explanation is that women are more sensitized to their health status and thus tend to report more conditions (Additional file 1: Table S2), but this should not influence MM defined according to objectively measured data, although the prevalence of objectively assessed psychiatric diseases was higher in women than in men (Additional file 1: Table S2). Overall, our results indicate that MM is more prevalent in Swiss women than in Swiss men, and that this difference is independent from other demographic, clinical or socio-economic characteristics.
Prevalence of MM increased considerably with age, a finding also reported in the literature [2,12,14,23]. This was mainly due to the increase in the overall number of participants with at least one condition (Additional file 2: Figures S4-S6). Indeed, age is related to functional decline and to an increase in the number of morbid conditions. Interestingly, the difference in the prevalence of MM using self-reported and measured data tended narrow with increasing age. This narrowing could be explained by a better awareness or a better diagnosis of the diseases; indeed, it has been shown that a significant percentage of cardiovascular risk factors are undiagnosed [20,21,25]. Overall, our results confirm that age is a strong determinant of MM, and that MM is prevalent even among young adults, a finding also reported elsewhere [26]. More importantly, our results suggest that, in developed countries, the total number of patients with MM will considerably increase in the forthcoming years because of the ageing population, with considerable impact on health care costs [27].
In agreement with other studies [28,29], prevalence of MM was higher among obese participants. The considerable increase of MM as defined by FCI among obese participants is easily explained by the fact that obesity is a criterion for MM according to FCI. Overall, our results reflect the clustering of risk factors and morbidities among obese subjects [29,30], and in future studies it will be of interest the joint trends in obesity and MM, namely in the younger population. Finally, the positive association between abdominal obesity and MM initially found on bivariate analysis became non-significant after multivariate analysis.
Being a former or current smoker was related to a higher prevalence of MM, a finding also reported previously [26]. As for obesity, the most likely explanation is the tobacco-induced increase of multiple morbid conditions. Hence and again, early smoking cessation should be offered to all current smokers in order to decrease their risk of disease and, consequently, of MM.
Low socio economic status, defined by a low education or by receiving social help, was positively associated with MM, although the association between educational level and MM was significant for the FCI definition only. Our results are in agreement with several studies [2,12,31,32] but not with another [33]. Possible explanations include the deleterious effect of working environment together with inadequate health behaviours among low SES groups. Overall, our results suggest that preventive measures should be directed to low SES groups, but such measures are different to implement [34] and their effects are controversial [35].
Swiss citizenship was positively associated with MM. This finding was somewhat unexpected as in Switzerland access to health care is available for all. Further, to our knowledge, no study ever focused on MM in migrants. Thus, reasons for this difference are not straightforward and can only be speculated. One possible explanation would be a higher use of the healthcare system by Swiss nationals, which would increase the likelihood of detecting diseases. Another possible explanation would be better health behaviours of migrants relatively to Swiss nationals [36], but future studies are needed to better assess this point.
Strengths and limitations
This study has several strengths. It was conducted on an apparently healthy, population-based sample. It also collected data on self-reported and measured morbidities, allowing the comparison of the two data collection methods.
This study has also several limitations. The participation rate (41%) was relatively low, although in line or event higher than other epidemiological studies [37,38]. Hence, a selection bias cannot be ruled out, subjects presenting with morbidities being more prone to refuse to participate. Similarly, some participants might have forgotten to report some diseases or incorrectly reported them. Still, this would lead to an underestimation of the true prevalence of MM within the target population; thus, we believe that our prevalence estimates, although relatively high, are even though rather conservative. Due to its cross-sectional design, no association could be made between MM severity and quality of life or mortality. The ongoing follow-up of the CoLaus cohort will enable assessing the trends in the prevalence of MM and the associations between the different definitions of MM and mortality. Finally, in Switzerland, matching of medical electronic records with information from surveys is not allowed. Thus, it was not possible to confirm the statements of the participants from electronic records. Still, several studies have shown a discrepancy between data collected by surveys and extracted from electronic health records [18,19]: depending on the disease of interest, the prevalence obtained from electronic records could be similar, higher or lower than the prevalence reported in the survey, with further variations according to age and sex [19].